03 Rich metadata
An identifier is an address. The record is everything else.
doi.org/10.····/····
Every object we publish gets a DOI, a permanent address. An address tells you where something is — not what it is, who made it, who paid for it or what it belongs with. That is the work of the metadata record, and it is where we put our effort.
A One object, drawn in plan
A dataset and the eight records it opens onto.
What each door means
IsNewVersionOf— its earlier release (DataCite DOI)The earlier release stays citable, and anyone who finds it is told a newer one exists.
IsSupplementTo— the article it supports (Crossref DOI)This data is the evidence behind a published article; readers of either can reach the other.
IsDocumentedBy— the paper that describes it (Crossref or DataCite DOI)A peer-reviewed description of how the data were made and how to use them.
IsReviewedBy— our quality-review report (DOI)What was checked, by whom, and what changed as a result — published, not hidden.
IsSourceOf— a model trained on it (DataCite DOI)Models built from the data point back to it, so credit and licence terms travel with them.
IsCompiledBy— the code that produced it (DataCite DOI)The exact software release used to build the dataset, so the work can be repeated.
References— what it cites (DOIs where they exist)Every source the data rely on, each one linked rather than typed as text.
IsDerivedFrom— the data it was built from (DOI or other identifier)Where the raw material came from — a survey, an archive, an instrument run.
B The promise
What every Metamahal record carries.
Every field that applies. Not only the fields needed to register the identifier, but every recommended and optional field the agency offers for that kind of object.
People and organisations identified, not just named. ORCID iDs for creators and contributors, ROR IDs for organisations, funder identifiers and award numbers for grants.
References as links. The full reference list deposited with the record, with a DOI for every work that has one.
Links between the parts. Article, data, code, model, review and earlier versions joined by named relations. We record them from our side and ask the other publisher or repository to record the reverse.
Reuse terms a machine can read. A licence URL on every record.
More than one language. Title, abstract and keywords in the language of the work and in English, and in further languages when you give them to us.
Changes in the open. New versions get their own identifier and point back; corrections and retractions are linked to the original, never silently replaced.
The record is yours, and everyone’s. You receive a copy of the full record in the agency’s own format, and we release our metadata under CC0.
C Measured, not claimed
Eleven elements Crossref checks for every member.
For records registered with Crossref, its Participation Reports show anyone how many of a member’s records carry each of these elements. They are the elements we deliver for every object they apply to:
- References
- Abstracts
- ORCID iDs
- Affiliations
- ROR IDs
- Funder Registry IDs
- Funding award numbers
- Crossmark enabled
- Text mining URLs
- License URLs
- Similarity Check URLs
Element names as Crossref writes them. Some depend on the services a member uses, such as Crossmark and Similarity Check.
For records registered with DataCite, the schema sets six properties as mandatory, six as recommended and eight as optional; its metadata working group “strongly urge the inclusion of metadata identified as Recommended”. We treat all twenty as the standard.
D Languages
A record in one language reaches readers of one language.
We record titles, abstracts and keywords in the language of the work and in English, and in further languages whenever authors provide them. Each is tagged with its language, so search services can show the right one.
When a work is itself translated, the translation and the original point to each other with the relation types the agencies define for it: isTranslationOf and hasTranslation at Crossref, IsTranslationOf and HasTranslation at DataCite, where translators can also be credited in their own role.